davideisinger.com

My personal website
Log | Files | Refs | README

www-programmablemutter-com-8wp6z1.txt (23165B)


      1 [1][https]
      2 
      3 [2]Programmable Mutter
      4 
      5 Subscribe
      6 Sign in
      7 Share this post
      8 [https]
      9 
     10 What OpenAI shares with Scientology
     11 
     12 www.programmablemutter.com
     13 Copy link
     14 Facebook
     15 Email
     16 Note
     17 Other
     18 
     19 What OpenAI shares with Scientology
     20 
     21 Strange beliefs, fights over money and bad science fiction
     22 
     23 [12][https]
     24 [13]Henry Farrell
     25 Nov 20, 2023
     26 73
     27 Share this post
     28 [https]
     29 
     30 What OpenAI shares with Scientology
     31 
     32 www.programmablemutter.com
     33 Copy link
     34 Facebook
     35 Email
     36 Note
     37 Other
     38 [20]
     39 17
     40 [21]
     41 Share
     42 [22]
     43 [https]
     44 
     45 When Sam Altman was ousted as CEO of OpenAI, some hinted that lurid depravities
     46 lay behind his downfall. Surely, OpenAI’s board wouldn’t have toppled him if
     47 there weren’t some sordid story about to hit the headlines? But the [23]
     48 reporting all seems to be saying that it was God, not Sex, that lay behind
     49 Altman’s downfall. And Money, that third great driver of human behavior, seems
     50 to have driven his attempted return and his [24]new job at Microsoft, which is
     51 OpenAI’s biggest investor by far.
     52 
     53 As the NYT describes the people who pushed Altman out:
     54 
     55 Thanks for reading Programmable Mutter! Subscribe for free to receive new
     56 posts. And if you want to support my work, [25]buy my and Abe Newman’s new
     57 book, [26]Underground Empire, and sing its praises (so long as you actually
     58 liked it), on Amazon, Goodreads, social media and everywhere else that people
     59 find out about good books.
     60 
     61 [35][                    ]
     62 Subscribe
     63     Ms. McCauley and Ms. Toner [HF - two board members] have ties to the
     64     Rationalist and Effective Altruist movements, a community that is deeply
     65     concerned that A.I. could one day destroy humanity. Today’s A.I. technology
     66     cannot destroy humanity. But this community believes that as the technology
     67     grows increasingly powerful, these dangers will arise.
     68 
     69 McCauley and Toner reportedly worried that Altman was pushing too hard, too
     70 quickly for new and potentially dangerous forms of AI (similar fears led some
     71 OpenAI people to bail out and found a competitor, Anthropic, a couple of years
     72 ago). The FT’s reporting [37]confirms that the fight was over how quickly to
     73 commercialize AI
     74 
     75 The back-story to all of this is actually much weirder than the average sex
     76 scandal. The field of AI (in particular, its debates around Large Language
     77 Models (LLMs) like OpenAI’s GPT-4) is profoundly shaped by cultish debates
     78 among people with some very strange beliefs.
     79 
     80 As LLMs have become increasingly powerful, theological arguments have begun to
     81 mix it up with the profit motive. That explains why OpenAI has such an unusual
     82 corporate form - it is a non-profit, with a for-profit structure retrofitted on
     83 top, sweatily entangled with a profit-maximizing corporation (Microsoft). It
     84 also plausibly explains why these tensions have exploded into the open.
     85 
     86 ********
     87 
     88 I joked on Bluesky that the OpenAI saga was as if “the 1990s browser wars were
     89 being waged by rival factions of Dianetics striving to control the future.”
     90 Dianetics - for those who don’t obsess on the underbelly of American
     91 intellectual history - was the 1.0 version of L. Ron Hubbard’s Scientology.
     92 Hubbard [38]hatched it in collaboration with the science fiction editor John W.
     93 Campbell (who had a major science fiction award named after him until 2019,
     94 when his racism finally caught up with his reputation).
     95 
     96 The AI safety debate too is an unintended consequence of genre fiction. In
     97 1987, multiple-Hugo award winning science-fiction critic Dave Langford [39]
     98 began a discussion of the “newish” genre of cyberpunk with a complaint about an
     99 older genre of story on information technology, in which “the ultimate computer
    100 is turned on and asked the ultimate question, and replies `Yes, now there is a
    101 God!'
    102 
    103 However, the cliche didn’t go away. Instead, it cross-bred with cyberpunk to
    104 produce some quite surprising progeny. The midwife was the writer Vernor Vinge,
    105 who proposed a revised meaning for “singularity.” This was a term already
    106 familiar to science fiction readers as the place inside a black hole where the
    107 ordinary predictions of physics broke down. Vinge suggested that we would soon
    108 likely create true AI, which would be far better at thinking than baseline
    109 humans, and would change the world in an accelerating process, creating a
    110 historical [40]singularity, after which the future of the human species would
    111 be radically unpredictable.
    112 
    113 These ideas were turned into novels by Vinge himself, including A Fire Upon the
    114 Deep (fun!) and Rainbow’s End (weak!). Other SF writers like Charles Stross
    115 wrote novels about humans doing their best to co-exist with “weakly godlike”
    116 machine intelligence (also fun!). Others who had no notable talent for writing,
    117 like the futurist Ray Kurzweil, tried to turn the Singularity into the
    118 foundation stone of a new account of human progress. I still possess a
    119 mostly-unread copy of Kurzweil’s mostly-unreadable magnum opus, The Singularity
    120 is Near, which was distributed en masse to bloggers like meself in an early
    121 2000s marketing campaign. If I dug hard enough in my archives, I might even be
    122 able to find the message from a publicity flack expressing disappointment that
    123 I hadn’t written about the book after they sent it. All this speculation had a
    124 strong flavor of end-of-days. As the Scots science fiction writer, Ken MacLeod
    125 memorably put it, the Singularity was the “Rapture of the Nerds.” Ken, being
    126 the [41]offspring of a Free Presbyterian preacher, knows a millenarian religion
    127 when he sees it: Kurzweil’s doorstopper should really have been titled The
    128 Singularity is Nigh.
    129 
    130 Science fiction was the gateway drug, but it can’t really be blamed for
    131 everything that happened later. Faith in the Singularity has roughly the same
    132 relationship to SF as UFO-cultism. A small minority of SF writers are true
    133 believers; most are hearty skeptics, but recognize that superhuman machine
    134 intelligences are (a) possible) and (b) an extremely handy engine of plot. But
    135 the combination of cultish Singularity beliefs and science fiction has
    136 influenced a lot of external readers, who don’t distinguish sharply between the
    137 religious and fictive elements, but mix and meld them to come up with strange
    138 new hybrids.
    139 
    140 Just such a syncretic religion provides the final part of the back-story to the
    141 OpenAI crisis. In the 2010s, ideas about the Singularity cross-fertilized with
    142 notions about Bayesian reasoning and some really terrible fanfic to create the
    143 online “rationalist” movement mentioned in the NYT.
    144 
    145 I’ve never read a text on rationalism, whether by true believers, by
    146 hangers-on, or by bitter enemies (often erstwhile true believers), that really
    147 gets the totality of what you see if you dive into its core texts and
    148 apocrypha. And I won’t even try to provide one here. It is some Very Weird Shit
    149 and there is really great religious sociology to be written about it. The
    150 fights around [42]Roko’s Basilisk are perhaps the best known example of
    151 rationalism in action outside the community, and give you some flavor of the
    152 style of debate. But the very short version is that [43]Eliezer Yudkowsky, and
    153 his multitudes of online fans embarked on a massive collective intellectual
    154 project, which can reasonably be described as resurrecting David Langford’s
    155 hoary 1980s SF cliche, and treating it as the most urgent dilemma facing human
    156 beings today. We are about to create God. What comes next? Add Bayes’ Theorem
    157 to Vinge’s core ideas, sez rationalism, and you’ll likely find the answer.
    158 
    159 The consequences are what you might expect when a crowd of bright but rather
    160 naive (and occasionally creepy) computer science and adjacent people try to
    161 re-invent theology from first principles, to model what human-created gods
    162 might do, and how they ought be constrained. They include the following,
    163 non-comprehensive list: all sorts of strange mental exercises, postulated
    164 superhuman entities benign and malign and how to think about them; the jumbling
    165 of parts from fan-fiction, computer science, home-brewed philosophy and ARGs to
    166 create grotesque and interesting intellectual chimeras; Nick Bostrom, and a
    167 crew of very well funded philosophers; Effective Altruism, whose fancier
    168 adherents often prefer not to acknowledge the approach’s somewhat disreputable
    169 origins.
    170 
    171 All this would be sociologically fascinating, but of little real world
    172 consequence, if it hadn’t profoundly influenced the founders of the
    173 organizations pushing AI forward. These luminaries think about the technologies
    174 that they were creating in terms that they have borrowed wholesale from the
    175 Yudkowsky extended universe. The risks and rewards of AI are seen as largely
    176 commensurate with the risks and rewards of creating superhuman intelligences,
    177 modeling how they might behave, and ensuring that we end up in a Good
    178 Singularity, where AIs do not destroy or enslave humanity as a species, rather
    179 than a bad one.
    180 
    181 Even if rationalism’s answers are uncompelling, it asks interesting questions
    182 that might have real human importance. However, it is at best unclear that
    183 theoretical debates about immantenizing the eschaton tell us very much about
    184 actually-existing “AI,” a family of important and sometimes very powerful
    185 statistical techniques, which are being applied today, with emphatically
    186 non-theoretical risks and benefits.
    187 
    188 Ah, well, nevertheless. The rationalist agenda has demonstrably shaped the
    189 questions around which the big AI ‘debates’ regularly revolve, as [44]
    190 demonstrated by the Rishi Sunak/Sam Altman/Elon Musk love-fest “AI Summit” in
    191 London a few weeks ago.
    192 
    193 We are on a very strange timeline. My laboured Dianetics/Scientology joke can
    194 be turned into an interesting hypothetical. It actually turns out (I only
    195 stumbled across this recently) that Claude Shannon, the creator of information
    196 theory (and, by extension, the computer revolution) was an [45]L. Ron Hubbard
    197 fan in later life. In our continuum, this didn’t affect his theories: he had
    198 already done his major work. Imagine, however, a parallel universe, where
    199 Shannon’s science and standom had become intertwined and wildly influential, so
    200 that debates in information science obsessed over whether we could eliminate
    201 the noise of our [46]engrams, and isolate the signal of our True Selves,
    202 allowing us all to become [47]Operating Thetans. Then reflect on how your
    203 imagination doesn’t have to work nearly as hard as it ought to. A similarly
    204 noxious blend of garbage ideas and actual science is the foundation stone of
    205 the Grand AI Risk Debates that are happening today.
    206 
    207 To be clear - not everyone working on existential AI risk (or ‘x risk’ as it is
    208 usually summarized) is a true believer in Strong Eliezer Rationalism. Most,
    209 very probably, are not. But you don’t need all that many true believers to keep
    210 the machine running. At least, that is how I interpret this [48]Shazeda Ahmed
    211 essay, which describes how some core precepts of a very strange set of beliefs
    212 have become normalized as the background assumptions for thinking about the
    213 promise and problems of AI. Even if you, as an AI risk person, don’t buy the
    214 full intellectual package, you find yourself looking for work in a field where
    215 the funding, the incentives, and the organizational structures mostly point in
    216 a single direction (NB - this is my jaundiced interpretation, not hers).
    217 
    218 ********
    219 
    220 There are two crucial differences between today’s AI cult and golden age
    221 Scientology. The first was already mentioned in passing. Machine learning
    222 works, and has some very important real life uses. [49]E-meters don’t work and
    223 are useless for any purpose other than fleecing punters.
    224 
    225 The second (which is closely related) is that Scientology’s ideology and
    226 money-hustle reinforce each other. The more that you buy into stories about the
    227 evils of mainstream psychology, the baggage of engrams that is preventing you
    228 from reaching your true potential and so on and so on, the more you want to
    229 spend on Scientology counselling. In AI, in contrast, God and Money have a
    230 rather more tentative relationship. If you are profoundly worried about the
    231 risks of AI, should you be unleashing it on the world for profit? That tension
    232 helps explain the fight that has just broken out into the open.
    233 
    234 
    235 It’s easy to forget that OpenAI was founded as an explicitly non-commercial
    236 entity, the better to balance the rewards and the risks of these new
    237 technologies. To quote from its [50]initial manifesto:
    238 
    239     It’s hard to fathom how much human-level AI could benefit society, and it’s
    240     equally hard to imagine how much it could damage society if built or
    241     used incorrectly. Because of AI’s surprising history, it’s hard to predict
    242     when human-level AI might come within reach. When it does, it’ll be
    243     important to have a leading research institution which can prioritize a
    244     good outcome for all over its own self-interest.
    245 
    246     We’re hoping to grow OpenAI into such an institution. As a non-profit, our
    247     aim is to build value for everyone rather than shareholders. Researchers
    248     will be strongly encouraged to publish their work, whether as papers, blog
    249     posts, or code, and our patents (if any) will be shared with the world.
    250     We’ll freely collaborate with others across many institutions and expect to
    251     work with companies to research and deploy new technologies.
    252 
    253 That … isn’t quite how it worked out. The Sam Altman justification for
    254 deviation from this vision, laid out in various interviews, is that it turned
    255 out to just be too damned expensive to train the models as they grew bigger,
    256 and bigger and bigger. This necessitated the creation of an add-on structure,
    257 which would sidle into profitable activity. It also required massive cash
    258 infusions from Microsoft (reportedly in [51]the range of $13 billion), which
    259 also has an exclusive license to OpenAI’s most recent LLM, GPT-4. Microsoft, it
    260 should be noted, is not in the business of prioritizing “a good outcome for all
    261 over its own self-interest.” It looks instead, to invest its resources along
    262 the very best Friedmanite principles, so as to create whopping returns for
    263 shareholders. And $13 billion is a lot of invested resources.
    264 
    265 This, very plausibly explains the current crisis. OpenAI’s governance
    266 arrangements are shaped by the fact that it was a non-profit until relatively
    267 recently. The board is a non-profit board. The two members already mentioned,
    268 McCauley and Toner, are not the kind of people you would expect to see making
    269 the big decisions for a major commercial entity. They plausibly represent the
    270 older rationalist vision of what OpenAI was supposed to do, and the risks that
    271 it was supposed to avert.
    272 
    273 But as OpenAI’s ambitions have grown, that vision has been watered down in
    274 favor of making money. I’ve heard that there were a lot of people in the AI
    275 community who were really unhappy with OpenAI’s initial decision to let GPT
    276 rip. That spurred the race for commercial domination of AI which has shaped
    277 pretty well everything that has happened since, leading to model after model
    278 being launched, and to hell with the consequences. People like Altman still
    279 talk about the dangers of AGI. But their organizations and businesses keep
    280 releasing more, and more powerful systems, which can be, and are being, used in
    281 all sorts of unanticipated ways, for good and for ill.
    282 
    283 It would perhaps be too cynical to say that AGI existential risk rhetoric has
    284 become a cynical hustle, intended to redirect the attentions of regulators
    285 toward possibly imaginary future risks in the future, and away from problematic
    286 but profitable activities that are happening right now. Human beings have an
    287 enormous capacity to fervently believe in things that it is in their
    288 self-interest to believe, and to update those beliefs as the interests change
    289 or become clearer. I wouldn’t be surprised at all if Altman sincerely thinks
    290 that he is still acting for the good of humankind (there are certainly enough
    291 people assuring him that he is). But it isn’t surprising either that the true
    292 believers are revolting, as Altman stretches their ideology ever further and
    293 thinner to facilitate raking in the benjamins.
    294 
    295 The OpenAI saga is a fight between God and Money; between a quite peculiar
    296 quasi-religious movement, and a quite ordinary desire to make cold hard cash.
    297 You should probably be putting your bets on Money prevailing in whatever
    298 strange arrangement of forces is happening as Altman is beamed up into the
    299 Microsoft mothership. But we might not be all that better off in this
    300 particular case if the forces of God were to prevail, and the rationalists who
    301 toppled Altman were to win a surprising victory. They want to slow down AI,
    302 which is good, but for all sorts of weird reasons, which are unlikely to
    303 provide good solutions for the actual problems that AI generates. The important
    304 questions about AI are the ones that neither God or [52]Mammon has particularly
    305 good answers for - but that’s a topic for future posts.
    306 
    307 Thanks for reading Programmable Mutter! Subscribe for free to receive new
    308 posts. And if you want to support my work, [53]buy my and Abe Newman’s new
    309 book, [54]Underground Empire, and sing its praises (as long as you actually
    310 liked it) on Amazon, Goodreads, social media and everywhere else that people
    311 find out about good books.
    312 
    313 [63][                    ]
    314 Subscribe
    315 73
    316 Share this post
    317 [https]
    318 
    319 What OpenAI shares with Scientology
    320 
    321 www.programmablemutter.com
    322 Copy link
    323 Facebook
    324 Email
    325 Note
    326 Other
    327 [71]
    328 17
    329 [72]
    330 Share
    331 17 Comments
    332 [https]
    333 [                    ]
    334 Share this discussion
    335 [https]
    336 
    337 What OpenAI shares with Scientology
    338 
    339 www.programmablemutter.com
    340 Copy link
    341 Facebook
    342 Email
    343 Note
    344 Other
    345 
    346         [81]
    347         Tarik Najeddine
    348         [82]Writes Factual Dispatch
    349         [83]Nov 20, 2023
    350 
    351 [80]    ChatGPT is just Zapp Brannigan or a McKinsey consultant. A veneer of
    352 [https] confidence and a person to blame when the executive "needs" to make a
    353         hard decision. You previously blamed the Bain consultants when you
    354         offshored a factory, now you blame AI.
    355 
    356         Expand full comment
    357         Reply
    358         Share
    359 
    360         [86]
    361         Gerben Wierda
    362         [87]Nov 21, 2023·edited Nov 21, 2023
    363 
    364         Came here via Dave Karpf's link. Beautiful stuff, and "The Singularity
    365         is Nigh" made me laugh out loud.
    366 
    367         The psychological and sociological/cultural side of the current
    368         GPT-fever is indeed far more important and telling than the technical
    369         reality. Short summary: quantity has its own certain quality, but the
    370         systems may be impressive, we humans are impressionable.
    371 
    372         Recently, Sam Altman received a Hawking Fellowship for the OpenAI Team
    373 [85]    and he spoke for a few minutes followed by a Q&A (available on
    374 [https] YouTube). In that session he was asked what are important qualities for
    375         'founders' of these innovative tech firms. He answered that founders
    376         should have ‘deeply held convictions’ that are stable without a lot of
    377         ‘positive external reinforcement’, ‘obsession’ with a problem, and a
    378         ‘super powerful internal drive’. They needed to be an 'evangelist'. The
    379         link with religion shows here too. ([88]https://
    380         erikjlarson.substack.com/p/gerben-wierda-on-chatgpt-altman-and). TED
    381         just released Ilya Sutskever’s talk and you see it there too. We have
    382         strong believers turned evangelists and we have a world of disciples
    383         and followers. It is indeed a very good analogy.
    384 
    385         Expand full comment
    386         Reply
    387         Share
    388 
    389 [90]15 more comments...
    390 Top
    391 New
    392 Community
    393 
    394 No posts
    395 
    396 Ready for more?
    397 
    398 [99][                    ]
    399 Subscribe
    400 © 2024 Henry Farrell
    401 [101]Privacy ∙ [102]Terms ∙ [103]Collection notice
    402 [104] Start Writing[105]Get the app
    403 [106]Substack is the home for great writing
    404 This site requires JavaScript to run correctly. Please [107]turn on JavaScript
    405 or unblock scripts
    406 
    407 References:
    408 
    409 [1] https://www.programmablemutter.com/
    410 [2] https://www.programmablemutter.com/
    411 [12] https://substack.com/profile/557668-henry-farrell
    412 [13] https://substack.com/@henryfarrell
    413 [20] https://www.programmablemutter.com/p/look-at-scientology-to-understand/comments
    414 [21] javascript:void(0)
    415 [22] https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F555fe47f-ac07-4614-b78b-5d269fde7539_1024x1024.webp
    416 [23] https://www.nytimes.com/2023/11/18/technology/open-ai-sam-altman-what-happened.html
    417 [24] https://www.ft.com/content/54e36c93-08e5-4a9e-bda6-af673c3e9bb5
    418 [25] https://amzn.to/3PbIyqX
    419 [26] https://amzn.to/3PbIyqX
    420 [37] https://www.ft.com/content/54e36c93-08e5-4a9e-bda6-af673c3e9bb5
    421 [38] https://longreads.com/2017/02/01/xenus-paradox-the-fiction-of-l-ron-hubbard/
    422 [39] https://ansible.uk/ai/pcwplus/pcwp1987.html
    423 [40] https://edoras.sdsu.edu/~vinge/misc/singularity.html
    424 [41] https://www.heraldscotland.com/life_style/arts_ents/14479010.science-fiction-writer-ken-macleod-free-presbyterian-childhood-time-communist-party-member-future-humanity/
    425 [42] https://www.lesswrong.com/tag/rokos-basilisk
    426 [43] https://en.wikipedia.org/wiki/Eliezer_Yudkowsky
    427 [44] https://www.politico.eu/article/rishi-sunak-artificial-intelligence-pivot-safety-summit-united-kingdom-silicon-valley-effective-altruism/
    428 [45] https://longreads.com/2018/10/23/the-dawn-of-dianetics-l-ron-hubbard-john-w-campbell-and-the-origins-of-scientology/
    429 [46] https://en.wikipedia.org/wiki/Engram_(Dianetics)
    430 [47] https://en.wikipedia.org/wiki/Operating_Thetan
    431 [48] https://crookedtimber.org/2023/11/16/from-algorithmic-monoculture-to-epistemic-monoculture-understanding-the-rise-of-ai-safety/
    432 [49] https://en.wikipedia.org/wiki/E-meter
    433 [50] https://openai.com/blog/introducing-openai
    434 [51] https://www.bloomberg.com/news/newsletters/2023-06-15/how-chatgpt-openai-made-microsoft-an-ai-tech-giant-big-take
    435 [52] https://www.newadvent.org/cathen/09580b.htm
    436 [53] https://amzn.to/3PbIyqX
    437 [54] https://amzn.to/3PbIyqX
    438 [71] https://www.programmablemutter.com/p/look-at-scientology-to-understand/comments
    439 [72] javascript:void(0)
    440 [80] https://substack.com/profile/1263175-tarik-najeddine
    441 [81] https://substack.com/profile/1263175-tarik-najeddine
    442 [82] https://factualdispatch.substack.com/?utm_source=substack&utm_medium=web&utm_content=comment_metadata
    443 [83] https://www.programmablemutter.com/p/look-at-scientology-to-understand/comment/43988738
    444 [85] https://substack.com/profile/23165546-gerben-wierda
    445 [86] https://substack.com/profile/23165546-gerben-wierda
    446 [87] https://www.programmablemutter.com/p/look-at-scientology-to-understand/comment/44033603
    447 [88] https://erikjlarson.substack.com/p/gerben-wierda-on-chatgpt-altman-and
    448 [90] https://www.programmablemutter.com/p/look-at-scientology-to-understand/comments
    449 [101] https://substack.com/privacy
    450 [102] https://substack.com/tos
    451 [103] https://substack.com/ccpa#personal-data-collected
    452 [104] https://substack.com/signup?utm_source=substack&utm_medium=web&utm_content=footer
    453 [105] https://substack.com/app/app-store-redirect?utm_campaign=app-marketing&utm_content=web-footer-button
    454 [106] https://substack.com/
    455 [107] https://enable-javascript.com/